R. Hoekema

Geometrical aspects of the interindividual variability of multilead ECG recordings

The ECG as measured from healthy subjects shows a considerable interindividual variability. This variability is caused by geometrical as well as by physiological factors. In this study, the relative contribution of the geometrical factors is estimated. In addition a method aimed at correcting for these factors is described. First, a measure (RV) for quantifying the overall variability is presented, and for healthy individuals its value is estimated as 0.52. Next, based on a simulation study using the individual (heart-lung-torso) geometry of 25 subjects, the variability caused by geometrical factors is estimated as 0.40, indicating that in healthy subjects the RV for healthy individuals resulting from electrophysiology is of the order of 0.33. In an evaluation of the correction procedure, applied to realistic, simulated body surface potentials, it is shown that RV caused by geometrical factors can be reduced from 0.40 to 0.06. When applying the correction procedure to measured ECG data no reduction of the RV value could be demonstrated. These results indicate that the involved procedure of the inverse computation of a cardiac equivalent source, at the present time, is of insufficient quality to cash in on the substantial reduction of RV values from 0.52 down to 0.33 that might be obtainable.

Geometrical aspects of the inter-individual variability of multilead ECG recordings

The ECG as measured from healthy subjects shows considerable inter-individual variability. This variability is caused by geometrical, as well as by physiological factors. A correction method for the variability due to the geometrical causes is proposed. In this method, the electrical heart activity is computed from the ECG in a realistic geometrical model. Subsequently, the electrical heart activity is transferred to a standard geometry and the resulting ECG is computed. In an evaluation study using simulated ECG data the inter-individual relative variability (RV) of the multilead ECG data was reduced from 0.40 to 0.06. Applied to measured ECG data so far no reduction could be obtained.

Lead system transformation of body surface map data

Multicenter application of body surface map data (multilead electrocardiographic [ECG] data) is hampered by the fact that the centers involved in body surface mapping use lead systems differing in lead placement as well as in the number of leads. In this study, the performance of two methods for converting multilead ECGs from one lead system to another is evaluated in their application to the major lead systems presently in use throughout the world. The first method is based on Laplacian interpolation, and the second method is derived from the correlations between the signals in an extensive lead system. Through analyzing the representation errors, it was found that, for lead systems incorporating over 60 leads, both methods work well, yielding errors comparable to interbeat differences in individuals. For lead systems incorporating fewer leads, the correlation method is to be preferred.

An elasticized net modeling the pericardium

In boundary element models used for computing pericardial potentials it is necessary to use individual geometry. This data is usually obtained from MR imaging. The triangulations describing the pericardium of individual subjects are not automatically node-to-node comparable, which hampers the analysis of inter-individual variability. This problem is now solved by a placing a stretchable, elasticized net mound the pericardial surface in such a way that anatomical landmarks defined on the pericardial net coincide with anatomical landmarks on the pericardial surface.

Normalization of the electrocardiogram by standardization of heart position and orientation

In healthy subjects, a large inter-individual variability in the ECG is present. Part of this variability is intrinsically due to physiological differences in the heart. Other causes are differences in the position and orientation of the heart relative to the chest. In the current study, research was done to reduce the variability through the latter cause. In healthy volunteers, 64-lead body surface maps (BSMs) were recorded and the inter-individual variability was determined. Pericardial potentials were calculated from the BSMs using an individual realistic torso and heart model for each subject. Then, the heart in the model was moved and rotated into a standard position and orientation. Finally, forward calculations were made using the new heart position, while keeping the pericardial potential distribution unaltered. This yielded new electrode potentials at the 64 BSM leads. The inter-individual variability of the corrected signals was calculated and compared with the original variability. The realistic models were based on magnetic resonance images of the torso, lungs and heart of four healthy volunteers. Using BSM and MRI data of 4 subjects, the authors have carried out the described procedure. The variability of the sets of BSM data was 0.28 mV. The pericardial potentials were calculated (forward simulation of the results yielded an error of 10%). The variability after shifting and rotating the hearts towards a standard position was 0.23 mV. It is concluded that a reduction of variability is indeed found when standardizing heart position and orientation. A larger number of subjects will be included in this study to substantiate this result.

Lead system transformation for pooling of body surface maps: an evaluation

Multi-center research on Body Surface Maps (multi-lead ECGs) is often hampered by the fact that there are many different lead systems in use, varying in lead placement as well as in the number of leads. In this study, two different methods for converting multi-lead ECGs from one lead system format to another are evaluated. The first method is based on Laplacian interpolation and the second method is based on the correlation between lead signals. It was found that the Laplacian transformation method works well for lead systems having more than 60 leads. For some smaller lead systems, the correlation method is to be preferred.